DETECTION OF ULTRA –WIDEBAND RADAR SIGNALS

SCATTERED FROM COMPLEX TARGETS

Igor I. Immoreev, Senior Member IEEE, Dmitry V. Fedotov

Moscow Aviation Institute. Gospitalny val, Home 5, block 18, apt 314. 105094, Moscow, Russia,
e-mail: immoreev@aha.ru, fedotovs@mtu-net.ru

The questions of detection of radar-tracking signals are discussed, which parameters are unknown. The method
for detection of radar signals which parameters are unknown is considered. The method is based on correlation
processing of signals received in the adjacent periods of sounding - the interleaved periodic correlation
processing (IPCP).

1. Introduction 2. Interleaved periodic correlation

Signals radiated by narrow band radar (the
instantaneous bandwidth ∆f ≤ 10% of the band
medium frequency) are harmonic or quasi-harmonic
oscillations. This makes an opportunity to perform
filter and correlation signal processing.
The signal shape differs from harmonic
oscillation more and more as the signal spectrum is
extended. In ultra wide band (UWB) radar (the
instantaneous bandwidth ∆f ≈ 100% of the band
medium frequency, radiated pulse width is about 1
ns) space signal duration is significantly less than
antenna size and target length. During the process of
target location a signal changes its shape many times
including the cases of signal scattering from target
bright points. As a result a returned signal is
transformed into the sequence of pulses with random
parameters. Target configuration and aspect angle
determine pulses arrangement. Pulses amplitude
depends on RCS of corresponding target bright point.
Pulses polarity depends on magnetic permeability of
the target material that scatters the signal. The initial
shape of the returned pulses is influenced by
resonance characteristics of target bright points.
So a narrow band signal changes its parameters
but holds its shape during the location process,
whereas a UWB signal changes its parameters and
initial shape as well. Such signal is often named as
“target image” because it carries knowledge of not
only target presence and target coordinates but also
target structure. Proper image processing makes
possible to recognize a target and to form radioimage
of target. At initial stage of location before target
recognition it is necessary to detect of target. It is not
advisable to use for UWB signal detection the
traditional methods such as optimum signal
processing by matched filtering or correlation with
the reference signal as the structure of UWB returned
signal is fully unknown.
processing
Methods for quasi-optimum processing of fully
unknown signals are well known. Some such
detectors are described in [1,2 and other]. But all
these detectors suffer from significant losses
compared to optimum detector.
The repetition period Tr is the only known
parameter in our case. This parameter may be used
for development of the optimum detector for UWB
signals, all other parameters we consider as
unknown. It uses as the reference signal a signal
received in the adjacent repetition period and delayed
at time interval Tr. So the received signal is not
compared to the reference signal as in a traditional
correlator but to the identical echo signal. So the
signal shape becomes the parameter that determines
the efficiency of such correlation detector. This
signal processing is named as interleaved periodic
correlation processing (IPCP). The schematic
diagram IPCP is shown in Fig. 1.
uecho(t)+un,2(t) U(Т)
× ∫
U thresh
uecho(t-Tr)+un,1(t)
Delay Тr
Fig. 1 Schematic diagram of interleaved periodic
correlation processing.
At this Figure: U(T) – voltage at correlator output;
uecho(t) – echo signal; un(t) – noise; Uthresh – threshold;
IPCP has three dissimilarities from the
conventional correlator. They are as follows:
a) the received signal is compared not with a radiated
signal but with a signal scattered by a target;
b) noises are fed to both correlator inputs (these transformation of normal noise distribution by the
noises are not correlated, as they are received in correlator components: a multiplier and an integrator.
different repetition periods); The distribution function Wm(y) for the
c) integration period T is determined not by the multiplication of two normally distributed random
radiated signal duration, but by the observation values y = u1 u2 describes the process at the
interval, that is, by the scattered signal duration. If a multiplier output when the echo signal is absent. If
physical target length is L the integration time is u1 and u2 are noncorrelated random values and
equal to T=(2L/c)-τ, where c is velocity of light and τ variances σ12 =σ22 the distribution is defined in [4] as:
is a duration of a radiated signal.  y 
1
Let us, first, determine the general laws of the W m (y ) = K 0  
2 
IPCP by the example of a signal scattered by a πσ 2
σ 
stationary target. In this case the signals at two where Ko(x) is the zero order Bessel‘s function of the
correlator inputs are noncorrelated noises having the 2nd kind for imaginary argument.
same normal distribution with zero mean values and The distribution function Wm(y) is illustrated in
a variance σ2 = No ∆f, where ∆f is UWB signal Fig.2.
bandwidth and N o is noise spectrum density. The
Neumann-Pearson criterion is used to estimate the
IPCP method efficiency.
The detection characteristics calculated for the
IPCP detector we compare with those for the
traditional optimum detector for fully known signal
Wm(y)
with fixed width received against white noise
background [2,3].
For the IPCP method the voltage at correlator
output is:
T
2
N 0 ∫0
U( Т) = u rec ( t ) u rec ( t − T r ) dt ;
Wn(y)
If the echo signal is absent, then:
Т T
2
U (T ) =
N 0 ∫0 ∫0 u n,1 (t) u n, 2 (t) dt
u n (t) u n (t − T r ) = Fig. 2: Distribution function for the product of two
normally distributed random values
Unlike the traditional correlator output the IPCP The normal distribution function Wn(u) area
correlator output is the multiplication of two normalized with Wm(y) is shown too. The
normally distributed and noncorrelated noises comparison of these two functions indicates that
received in two adjacent repetition periods. The multiplication of two normally distributed noises is
distribution function for this voltage Wo determines less dispersed relative to the mean value compared to
the threshold value required for given false alarm the initial normally distributed noise at the correlator
rate. In the IPCP correlator the threshold does not input. The reason is that bursts of the first noise
depend on the reference signal and has a constant signal are compensated by the low level of the other
value. noise signal. The coincidence of two burst is of low
If the echo signal is present in the received signal, probability. None the less this event may occur ant it
then: results in more extended “tails” of distribution
2 
T T
 function Wm(y) than those of distribution function
U( Т ) =  ∫ echo
u (t) u echo (t) dt + ∫ u echo (t) u n,1 (t) dt  + Wn(u). This fact is important for the IPCP correlator
N0  0 0  as the threshold level determined by the function
2  
T T
Wm(y) may be higher than the threshold level
+  ∫ u echo (t) u n,2 (t) dt + ∫ u n,1(t) u n,2 (t) dt  determined by the function Wn(u) for the same
N0  0 0 
values of false alarm rate.
As in the previous case the first integral
The multiplication of noises from the multiplier
determines nonzero mean value of the distribution
output is fed to the integrator. The type of
function at the correlator output W1. The second and
distribution function at integrator output depends on
the third integrals determine this function variance
the integration interval T. According to the central
when the echo signal is present. The fourth integral
limit theorem this distribution approach the normal
determines the distribution function type.
distribution under integration and its variance grows
To calculate the detection characteristics we find
proportionally to the interval T.
distribution functions Wo and W1 for the IPCP
correlator. For this purpose we consider the

Fig.3: Dependence of distribution function Wm(y) on
integration period
If the UWB signal is processes by the IPCP
correlator the integration interval is determined not
by the duration of a radiated signal as in the
traditional narrowband correlator, but by the real
target length and equal T=(2L/c)-τ. If UWB signal
duration τ = 1 ns the real targets can occupy from 10
to 100 range cells, the integration interval can vary
from 10τ to 100τ correspondingly. This time interval
may be too short for full normalization of the noise
multiplication distribution function Wm(y). So this
function will have an intermediate shape between the
noise multiplication distribution and the normal
distribution, approaching the normal distribution as
the integration interval is increasing. Fig.3 shows
this distribution in relation to the integration interval
T. T varies from 10τ to 80τ in the interval of 10τ.
The development of analytical expression for
distribution function at the integrator output presents
a real difficulty if the distribution of input signals
Fig. 4: Distribution function W0(u) and W(u) for
IPCP
differs from the normal distribution. So mathematical
simulation was used to calculate and plot the
distribution functions Wo and W1 at the IPCP
correlator output. The approximate view and mutual
positions of functions W1 and W0 at IPCP are shown
in Fig.4.
As it is rather difficult to have analytical expressions
for distribution functions W1 and W0 at the IPCP
outlet, we use mathematical modeling to plot
detection characteristics.
3. Detection characteristics
Fig.5,a and 5,b show the detection characteristics of
IPCP for a signal scattered by a stationary target for
two values of false alarm rate 10-2 and 10-4 (D -
probability of detection, q - ratio signal – noise).
These Figures also show the detection characteristics
Fig. 5: Detection curves for a signal scattered by a
stationary target at IPCP
1- IPCP for stationary target + Criterion processing;
2- Traditional correlator for fully known signal;
3- IPCP for stationary target;
4- Energy detector.
of a conventional correlator for fully known signal
and detection characteristics of the energy detector
for the same false alarm rates. In order to make the
comparison valid, the duration of a received signal is
taken equal to the duration a radiated signal (one
point target).
The analysis of results shows the following.
The IPCP detection characteristics approach the
conventional correlator detection characteristics for
high false alarm rates (10-2). The difference between
positions of these characteristic increases with
reducing the false alarm rates (10-4). This can be
explained by the long duration of the “tails” of the
distribution function Wm(y). In IPCP the given level
of false alarm rate can be maintained by setting the
threshold level higher than in the conventional
correlator. At the same time the detection
characteristics of IPCP are much better, than those of
the energy detector.
Fig.6 shows the dependence of detection
characteristics on the integration time T (determined
by a target length). The false alarm rate 10-4 and the
integration time T equal to 2τ, 10τ, and 20τ were
Fig. 6: Dependence of detection parameters of IPCP
on integration period
1- Traditional correlator for fully known signal for
T=2τ;
2- IPCP for a stationary target for T=20τ;
3- IPCP for a stationary target for T=10τ;
4- IPCP for a stationary target for T=2τ;
5- Energy detector for T=2τ.
taken for modeling. This Figure also shows the
detection characteristic for a conventional correlator.
It is shown from the picture that with increasing the
target length the IPCP detection characteristic for a
stationary target approaches more and more the
conventional correlator characteristic when it detects
the fully known signal. The reason for this is that
distribution function Wm(y) approaches the normal
distribution, while integrating noise samples.
4. Criterion processing of UWB signals
after IPCP
The IPCP detection characteristics can be improved
by additional criterion processing. The criterion
processing scheme memorizes the resolution cells in
which the output signals from the threshold scheme
are present. This operation is performed in several
pulse repetition periods Tr, after that the cells in
which the signals emerge repeatedly are determined.
Only the signals from resolution cells which
correspond the criterion selected (“two of two”, “two
of three”, “three of four”, and etc.) are passed
through the scheme. This results in great reduction in
false alarms at the processor output. But at the same
time the detection probability is also decreasing. It
should be mentioned that in practice the detection
probability is always higher than the false alarm rate.
So the detection probability is decreasing more
slowly than the false alarm rate. The criterion
processing can be effectively used when the low
level of false alarm rate is required (10-4 and less).
The schematic diagram of such criterion
processing is illustrated in Fig.7. The signal samples
Fig. 7: Schematic diagram of criterion processing
from the same range cells received in two repetition
periods are fed to the AND logical scheme. Fig. 5a
and Fig. 5b (dotted line) show the IPCP detection
characteristics after the criterion processing “two of
two”. The characteristics are calculated for the false
alarm rate indicated in the Figure.
They are located to the left of the standard
characteristics of the conventional correlator
calculated for the fully known signal. This is a result
of combining signals from two repetition periods in
OR scheme (in this case, it operates as a multiplier),
this procedure is identical to the accumulation.
5. Moving target
If a target is moving the problem concerned with its
passing from one resolution cell to another during
pulse repetition period emerges. We can solve this
problem by using a multi-channel scheme similar to
Doppler filtration system, which provides optimal
detection of moving targets. The similar multi-
channel scheme can be used for selection of optimal
integration time T while detecting target with various
physical lengths L. The losses resulted from the
multi-cannel configuration of the scheme can be
calculated using the conventional methods that are
valid for the similar multi-cannel digital Doppler
systems.
6. References
[1] Van der Spek G.A. Detection of a distributed target.
IEEE Trans., 1971, sept., AES-7, № 5, p. 922-931.
[2] Sosulin Yu.G. Theoretical Foundations for Radar
Detection and Radio Navigation: - Moscow: Radio and
Communication, 1992.
[3] Gutkin L.S. The Theory of Optimum Methods for
Radio Detection in Fluctuating Noise: - Moscow:
Soviet Radio, 1972
[4] Levin B.R. The Theory of Random Processes
and Its Implementation in Radio Engineering: -
Moscow: Soviet Radio, 1960
[5] Immoreev, I.; “Features of Signals Detecting in
the UWB Radars” in book «Ultrawideband
Radar Technology», Edited by James D. Taylor,
CRC Press, Boca Raton, London, New Work,
Washington, 2000.